A comparative study of multi-objective evolutionary algorithms to optimize the selection of investment portfolios with cardinality constraints

Abstract

We consider the problem of selecting investment components according to two partially opposed measures: the portfolio performance and its risk. We approach this within Markowitz's model, considering the case of mutual funds market in Europe until July 2010. Comparisons were made on three multi-objective evolutionary algorithms, namely NSGA-II, SPEA2 and IBEA. Two well-known performance measures are considered for this purpose: hypervolume and R 2 indicator. The comparative analysis also includes an assessment of the financial efficiency of the investment portfolio selected according to Sharpe's index, which is a measure of performance/risk. The experimental results hint at the superiority of the indicator-based evolutionary algorithm. © 2012 Springer-Verlag.